1 Introduction

1.1 Background and Motivation

Employers, military agencies, and individuals all have shown an increased interest in developing emotional resilience, emotional regulation, and attention regulation skills. The increased interest in these skills has helped underpin a surge in research on mindfulness-based interventions (MBIs). MBIs are frequently based on Eastern contemplative teachings, and are often associated with various forms of meditation. Westernized versions of MBIs, like Mindfulness-Based Stress Reduction (MBSR) [1], are usually adapted (e.g., secularized) to be more acceptable to other audiences and sometimes combined with other therapies to treat specific clinical diagnoses.

At the root of MBIs is the concept of mindfulness; Bishop describes the state of mindfulness as “A kind of nonelaborative, nonjudgmental, present-centered awareness in which each thought, feeling, or sensation that arises in the attentional field is acknowledged and accepted as it is [2].” By focusing on the present and not elaborating or attaching oneself to their thoughts, instances of negative thought loops are reduced. Instead of ruminating on the past (depression) or worrying about the future (anxiety), attention is placed on one’s current, embodied experience. The ability to focus on the present moment, recognize emotional stimuli, and choose how to respond (non-reactivity) is cultivated through regular practice. Practice may include mindful meditation, body scanning, and elements of yoga.

With consistent practice, it becomes easier to achieve a mindful state. However, Davidson argues that focusing on the state of mindfulness is problematic because states fade, so any intervention focused on states is not much different than pharmaceutical interventions [3]. Instead, we should focus on developing the trait of mindfulness, which is the generally tendency to be in mindful states. One of the main issues is, as Kiken and colleagues found [4], people vary significantly in their response to mindfulness interventions and ability to develop the trait of mindfulness. There are many personal, environmental, and logistical barriers to sustained practice that leads to trait mindfulness; we have discussed these barriers, including: difficulty tuning out external stimuli, difficulty participating in group sessions, inaccessibility of qualified instructors, in the past and refer you to [5, 6] for additional information on those barriers.

The barriers to practice as well as the general difficulty of mindfulness has opened up opportunities for HCI researchers to work on making MBIs more accessible. HCI researchers explore the use of smart phone apps, virtual reality environments (VREs), and biofeedback [7] to enhance MBI practice. The work we discuss today focuses on addressing one particular accessibility challenge associated with practicing mindfulness, directed attention fatigue [REF], through the use of virtual reality. Directed attention fatigue (DAF) is a concept from attention restoration theory; DAF is based on the assumption attention is a limited resource and that our modern environment drains that resource. The depletion of attentional resources makes mindfulness practice difficult; in some ways, it is analogous to exercising while tired. We will discuss attention restoration theory in the next section.

1.2 Conceptual Framework

Attention restoration theory (ART) posits that attention is a limited resource and that our modern environment, which includes the built environment and technologies designed to capture our attention, tax that limited resource [8, 9]. When that resource becomes depleted, people develop directed attention fatigue. Directed attention fatigue reduces one’s ability to focus and can also lead to other negative psychosocial outcomes, including irritability and decreased performance on mental tasks [10]. Longer term, Kuo and Sullivan [11] found that city residents with less access to green space report higher levels of mental fatigue, irritabiltiy and violence. Researchers have argued that attention is a key resource in executive functioning, and excessive fatigue may undermine attempts to practice mindfulness-based interventions because attentional control is a key element of those practices [12]. Mindfulness requires purposeful allocation of attentional resources; the ability to direct one’s attention is cultivated through practice but excessive fatigue will make starting MBIs difficult.

One potential treatment to DAF is exposure to nature [8]. According to ART, nature has inherently restorative effects on attention fatigue and emotional well-being. The mechanisms through which nature has these effects is unknown, but one theory is that humans evolved to function within natural environments, and thus may be better adapted to functioning within them. Although the mechanisms of nature exposure are unknown, researchers identified five factors of natural space that are associated with the perceived restorativeness of a space: away, fascination, coherence, compatibility, and extent. The factors are defined as follows:

Away: a sense of being separated (physically or mentally) from daily concerns or problems.

Fascination: in this case, soft fascination, where elements of the environment capture one’s attention.

Compatibility: whether the individual views the environment as suitable to their interests. For example, some people prefer ocean side views while others prefer secluded glades in a forest [13].

Extent-Scope: has two sub-factors, scope and coherence. Coherence is the extent to which elements of the environment fit together. In some variations of the scale, coherence is separated out as a fifth factor [14]. Scope refers to whether the space is of sufficient size and complexity to be engaging.

We were particularly interested in whether virtual reality could serve as a delivery mechanism for mindfulness instruction and whether the medium could replicate some of the purported benefits of nature exposure. Researchers are actively exploring whether exposure to nature themed environments, including digital representations, generate partial effects under the premise that most people have limited access to natural environments. Our general research question was: can a nature inspired virtual reality-based meditation environment create a restorative effect and improve meditation session quality?

The degree to which an individual perceives an environment as meeting the qualities of away, fascination, compatibility, and extent-scope influences whether the environment has restorative properties for that individual. ART states that a restorative environment will serve to replenish the attentional resources of the individual and prepare them physiologically (i.e., reduce arousal due to over-stimulation) for the mindfulness practice. Therefore, our first hypothesis is:

  • H1: Perceived restorativeness is correlated with reductions in arousal.

More specifically, we were focused on physiological arousal of the individuals. One well-understood measure of arousal is electrodermal activity (EDA). EDA is a measure of skin conductance, which is affected by sweat gland activity [15]. Sweat glads are controlled by the sympathetic nervous system, and are not under direct control of the individual and are therefore thought to be a more accurate measure of arousal.

There are two components of EDA: phasic and tonic activity. Tonic EDA, or skin conductance level (SCL) is the general background level of arousal. In contrast, phasic EDA, or skin conductance response (SCR) are event driven. Positive changes in SCL can indicate an increased level of arousal over time, while negative changes indicate a decreased level of arousal. SCRs are due to specific stimuli, with the relative frequency of events being correlated with sensitivity to environmental stimuli. Therefore, we further clarify our hypothesis as:

  • H1a: Perceived restorativeness is significantly positively correlated with a reduction in skin conductance level.

  • H1b: Perceived restorativeness is correlated with a reduction in the frequency of skin conductance responses.

Hypothetically, a restorative space should make it easier to engage in mindfulness practice. If the space distances the individual from their normal cares, is mildly fascinating but not overly stimulating, matches what they perceive to be a relaxing natural environment, large enough to feel as if they are immersed in a new place, and put together in a way that does not detract from the experience, then the individual’s arousal levels and mental fatigue will diminish. We hypothesize that a feeling of restorativeness and the related reduction of arousal will make it easier for individuals to practice mindfulness. Therefore, we hypothesize:

  • H2a: Meditation depth [16] is significantly positively correlated to perceived restorativeness.

  • H2b: Meditation depth is significantly positively correlated with a decrease in skin conductance level.

  • H2c: Meditation depth is significantly positively correlated to a decrease in skin conductance responses.

Virtual environments vary in quality and immersiveness; design choices can affect the emotional impact of the environments as well as the extent to which individuals feel as if they are transported to the virtual environment. The concept of presence has evolved to describe the psychological state of transportation and non-mediation [17, 18]. That is to say, presence is often associated with an individual’s sensation that they are interacting directly with objects or actors in the virtual environment, not with or through a media technology.

Our main question was whether we could simulate a natural environment well enough to generate a perception of restorativeness and if that perceived restorativeness would increase the perceived quality of the mindfulness session. Formally, our hypotheses related to presence were:

  • H3a: Perceived restorativeness is significantly positively correlated with presence.

  • H3b: Meditation depth is significantly positively correlated with presence.

We describe the experimental methods and data collection procedures related to the above hypotheses in the following section.

2 Methods

All experimental recruitment procedures and methods were approved by Syracuse University’s Institutional Review Board. Participants were recruited from the local Syracuse, New York USA community through flyers and email lists. There was a small monetary compensation incentive advertised in the recruitment flyers. Those who responded to the solicitation were scheduled to visit the lab at a time that was convenient for them. Participants were greeted at the lab at their scheduled time and after explaining the nature of the study, given a consent form. If the participants consented to participate in the study, they were asked to complete a pre-test survey. After the participant completed the pre-test measures, the technician informed them that they would be hooked up to several non-invasive sensors (see below) and have a virtual reality head-mounted device placed on their head. All participants were allowed to terminate participation but none chose to do so.

Our virtual reality stimulus was a nature-themed meditation app with features found to be correlated with inducing a sense of restorativeness. In particular, we designed an open space adjacent to a waterfall, which created a running water effect. The area was sunny, with visual effects of a gentle breeze. We also included bird song as that has been found to improve relaxation. The environmental volume levels were held constant for all participants.

Participants were given a minute of silence to acclimate to the virtual environment before the meditation track began. The track was recorded and edited in a professional studio to reduce extraneous, distracting noise associated with recordings done in an office space. The track itself was written and narrated by a certified Mindfulness-Based Stress Reduction instructor. The 10-minute mindfulness track centered on awareness of breath. The instructor also leveraged elements of the virtual environment as objects of focus.

When the track was complete, the technician removed the sensors and virtual reality headset from the subjects. Finally, the participants completed the post-test measures via an online survey. Finally, participants were reimbursed and thanked for their time.

2.1 Measures

Our pre-test measures included basic demographic information, meditation experience (hours practiced) and style if applicable, and meditation retreat experience. Meditation experience as well as prior retreat attendance has been shown to affect an individual’s overall mindfulness and manifests as physiological change in neural circuitry.

We outlined our measures in a previous paper [6] but include the information for specific measures discussed in this paper below for the reader’s convenience. The following self-report measures were used:

Five Facets of Mindfulness Questionnaire (FFMQ; [19]) [Pre-test]. FFMQ is a 39-item, five factor instrument designed to capture one’s mindful disposition towards daily life. The five factors are: observing, describing, acting with awareness, non-judging of inner experience, and non-reactivity to inner experience.

Positive and Negative Affect Scale – Short Form (PANAS-SF, [20]). [Pre- and Post- test] We used a modified version of the PANAS-SF, worded to capture the strength of their current positive and negative moods. Subjects completed both a pre- and post- test PANAS to capture changes in mood due to the mindfulness session.

Perceived Stress Scale (PSS, [21], 1983) [Pre-test]. PSS consists of ten items grouped into two dimensions – positively and negatively worded questions related to how one appraises the stressfulness of situations in their life.

Perceived Restorativeness Scale (PRS, [22]) [Post-test] PRS is a four-factor, 26-item scale intended to capture properties of natural environments that are expected to facilitate emotional and attentional rejuvenation. Items are rated on a scale of 1–5, factors are calculated as a mean of all items in that factor, and the Five Facets Score is a mean of the factors.

SUS Presence Questionnaire (SUSPQ, [23]) [Post-test] SUSPQ has six questions, focused on three presence indicators: sense of being there, extent to which the virtual environment becomes more “real” than reality, and the extent to which the virtual environment is thought of as a place visited [6]. Items are rated on a scale of 1–7 and presence is calculated as a mean of the items.

Meditation Depth Questionnaire (MEDI, [16]) [Post-test]. MDQ was Developed to measure the depth or intensity of a meditation experience while simultaneously being agnostic to the school of meditation. The thirty-item scale has five factors – hinderances, relaxation, personal self, transpersonal qualities, and transpersonal self. Items are rated on a scale of 1–5, factors are calculated as a mean of all items in that factor, and the Five Facets Score is a mean of the factors.

Unless stated otherwise, all data were collected with a BIOPAC MP-150 system. The following physiological signals were collected during the experiment:

Electrodermal Activity (EDA) for arousal [15]. The sampling rate was set at 2000 Hz. Raw data from the module pair is bandlimited from DC to 10 Hz for both channels. Two sensors were placed on the participant’s non-dominant hand.

Electrocardiogram (ECG) for heart rate variability for arousal [5]. The module pair has a fixed gain rate of 2,0000Db, with bandlimits from 1 Hz to 35 Hz. Two sensors were placed on the participant’s collar bone and a ground electrode on the right hip.

Respiration Sensor (RSP) for breath rate. Raw RSP data from the module pair is bandlimited from DC to 10 Hz, to provide for the measurement of relatively static respiratory conditions, such as cessation of breathing, up to the extremely rapid respiratory effort variations (up to 600 breaths per minute) associated with coughing or sneezing. An elastic band with a pressure sensor was wrapped around participant’s chest so that the pressure sensor was located at the inferior section of the participant’s sternum.

Function Near Infrared Spectroscopy (fNIRS) for hemodynamic patterns of the brain for assessing cognitive states [24, 25]. Captured with a Hitachi ETG-4000, sampling oxygenated and deoxygenated hemoglobin at 10 Hz. Probes were placed in 2, 3 \(\times \) 5 arrangements above both the right and left frontal cortex, which resulted in 24 channels of data.

3 Results

Although we collected a number of self-report and physiological measures, only a subset of the analysis will be discussed below. We provided the comprehensive list of measures as additional information for readers who may be interested or have suggestions about how to resolve some of the contradictory findings in our analysis (see below).

The associations between the following variables are discussed below, along with descriptive statistics for each variable: presence, perceived restorativeness, meditation depth, and electrodermal activity.

3.1 Demographics

Twenty-one participants (Female = 10) completed the experiment; the average age of participants was 27.9 years old. Over half (n = 11) reported some college, while five of the participants reported having a graduate degree. The range of meditation experience was substantial - mean hours practiced was 568 h, with the maximum at 10,000 h. However, most of our subjects had no experience (Table 1).

Table 1. Descriptive statistics: demographics
Full size table

3.2 Self-report Measures

All of the participants completed 100% of the self-report questions. We calculated the mean score for the factors in scales with sub-factors. Previous analysis indicated all scales had high internal consistency <REF>. Final scores for each variable were calculated using the mean of the scale’s factors, in accordance with published guidelines.

Our analysis of the presence scale deviated from published instructions outlined in [23], which suggested presence should be determined by the number of sixes (6) and sevens (7) reported by each participant. We found in previous analyses that the authors’ suggested approach was not informative when it came to analyzing associations between the variables. We will discuss this in greater detail later, but we attribute the lack of variance captured by \({<}ref{>}\)’s approach to the difficulty participants have in summarizing their entire experience with a single measure. Participants reported moderate levels of presence (M = 3.76/7, S.D. = 1.02). Ratings on the perceived restorativeness of the environment were clustered around 3.20–3.66 (out of 5) (see Table 2, except for coherence. The high coherence rating suggests that the design choices of the virtual environment fit together and were appropriate.

Table 2. Descriptive statistics: perceived restorativeness scale
Full size table

Table 3 contains the results of our descriptive analysis of participants’ reports on meditation depth and quality. Basic measures related to their ability to get into a meditative state indicate that the virtual environment was moderately supportive of their efforts (see Hindrances and Relaxation). The Transpersonal Qualities, Transpersonal Self, and Personal Self factors all had lower scores, which should be the case given that those factors describe states that are available to experienced meditators <ref>. It should be noted that hindrance items were reverse coded such that participants would rate the experience as being more hindering to their meditation session. Values presented here were reversed such that a higher score means fewer or lower levels of hindrances.

Table 3. Descriptive statistics: meditation depth
Full size table

3.3 Correlation Analysis

All correlations were calculated using R; unless stated otherwise, correlations refer to Pearson’s correlation.

For EDA analysis, we looked at change over time. For tonic EDA or skin conductance level (SCL), we analyzed mean EDA levels for two minutes during the fist two minutes, the halfway mark, and the last two minutes. All data were processed using the Python package Neurokit [26]. This follows the reasoning described in [15] and conducted in [27], that interventions can either raise or lower overall arousal levels. From those mean values, we calculated the change or delta between ending SCL and beginning SCL, with negative values indicating a reduction in arousal.

We also looked at the frequency and amplitude of skin conductance responses (SCRs). SCRs indicate how frequently an individual reacts to environmental triggers [15]. The amplitude of the responses indicate how strong those reactions are. We expected a reduction in the frequency and amplitude of SCRs between the first two minutes and last two minutes of the experience. We did see changes, but none were correlated with self-report measures (see Tables 4 and 5).

Table 4. Associations between factors of Meditation Depth (MEDI) and Electrodermal Activity. Delta refers to the difference between the last two minutes of the stimulus and the fist two minutes (*** < .001, ** < .01, < .05)
Full size table
Table 5. Associations between factors of Perceived Restorativeness (PRS) and Electrodermal Activity Delta refers to the difference between the last two minutes of the stimulus and the fist two minutes (*** < .001, ** < .01, < .05)
Full size table

Presence and Electrodermal Activity Presence and EDA.

Table 6. Associations between Presence and Electrodermal Activity (EDA) Delta refers to the difference between the last two minutes of the stimulus and the first two minutes (*** < .001, ** < .01, < .05)
Full size table

Perceived Restorativeness and Presence. We see moderate correlations between the factors of Away, Fascination, and Compatibility, as well as overall perceived restorativeness of the environment, and presence. Conceptually, there is some overlap between Away and presence, which would suggest that there should be a stronger correlation between the two variables. However, it appears as if participants are reporting that they felt partially transported to a new environment, but still away from their daily concerns. This could mean that being distracted by something that is slightly fascinating is sufficient enough to have a restorative effect.

Table 7. Relationship between presence and perceived restorativeness (*** < .001, ** < .01, < .05)
Full size table

Meditation Depth and Presence. Presence was moderately correlated with several factors of meditation depth. The associations between hindrances and presence suggests that elements of the virtual environment could have a distracting effect on the participant. One experienced meditator noted that the narrative was distracting and that it took awhile to get settled because they experienced a hard fascination versus soft fascination <ref>. That is, they spent too much time looking at the environment. The scores could also suggest that the virtual environment shields the participant from external stimuli.

Table 8 highlights the results of correlation analysis between presence and meditation depth (MEDQ). Presence was correlated with all factors of meditation depth except relaxation and transpersonal qualities. Presence was also correlated with the composite MEDI score.

Table 8. Relationship between Meditation Depth (MEDQ) and Presence (*** < .001, ** < .01, < .05)
Full size table

Perceived Restorativeness and Meditation Depth. We saw strong correlations between the perceived restorativeness of the environment and meditation depth (0.7) (see Table 9). In particular, a sense of being away was highly correlated with lower levels of hindrances and higher levels of relaxation. Extent-Scope did not have strong correlations with the meditation depth factors, and fascination and compatibility had only moderate correlations with meditation depth factors. Overall, we see low or moderate correlations between the perceived restorativeness of the environment and meditation depth factors associated with skilled practitioners. We do not consider this to be a drawback of the environment because skilled practitioners should not need a virtual environment to practice. The idealized environment is designed to help novices get past the first stages of mindfulness practice, when tuning out external distractions is a challenge.

Table 9. Relationship between Meditation Depth (MEDQ) and Perceived Restorativeness (*** < .001, ** < .01, < .05)
Full size table

4 Discussion

In contrast to previous studies, we found no relationship between changes in arousal, either measured as tonic electrodermal activity or frequency of skin conductance responses, over time and self-reported measures of perceived restorativeness. Therefore, we reject the first set of hypotheses that perceived restorativeness is positively correlated with a reduction in skin conductance level and that perceived restorativeness is significantly positively correlated with a reduction in the frequency of skin conductance responses.

We found a significant (p < .001) and strong correlation (0.71) between Meditation Depth and Perceived Restorativeness, as well as significant and strong correlations among the factors in each scale. Therefore, we accept hypotheses that meditation depth is significantly positively related to perceived restorativeness.

However, we found no significant associations between meditation depth and levels of arousal. Therefore, we reject the that meditation depth is positively related to a decrease in skin conductance level and that meditation depth is significantly positively related to a decrease in skin conductance responses.

We did find significant (p < .05) and moderate correlations between presence (r = 0.45) and perceived restorativeness (r = 0.52) as well as presence and meditation depth. Therefore, we accept the hypotheses that perceived restorativeness is positively correlated with presence and that meditation depth is positively correlated with presence.

We also found no associations between electrodermal activity and presence. The lack of correlations between EDA measures and self-report measures is perplexing as other researchers have found significant associations between the measures in similar experiments [13, 28]. We do not have a specific explanation, although the results could be due to our smaller sample size than Browning and colleagues. Andersen et al. [13] had a smaller sample size yet found significant effect of time and scene in changes in EDA. However, they intentionally induced stress prior to entering the virtual reality scene, which may have helped create a sufficiently large change to be statistically significant. Overall, we saw a mean reduction in tonic EDA with 81\(\%\) of participants exhibiting a decrease in tonic EDA over the experiment. The few outliers who did experience an increase in tonic EDA during the experiment may have skewed the results.

Based on the self-report measures, we do see evidence that a restorative virtual environment is positively associated with meditation depth and quality. The small sample size of our project made it difficult to conduct more complex analysis that would help us tease out the differences in perceived restorativeness and meditation depth. Anecdotally, we got feedback from our experienced meditators. Some were fascinated by the environment and needed more time to settle, while others noted that the pace of the track or small details distracted them from the meditation.

In the future, we would like to focus on providing more variety in terms of meditation environments with the expectation that people would find environments that they are more compatible with. We would also like to explore how meditation experience affects interaction with virtual nature environments as well as baseline mood. We did do preliminary analysis on changes in affect but found no statistically significant correlations, unlike <browning>. Finally, we could look to do a controlled study, analyzing whether virtual reality provides any benefit beyond a smart phone app in terms of meditation session quality.